Do we still need a full IDE in the age of AI agents?

i have been working on building an editor from the ground up, dropping all kinds of features that modern IDEs come with, including the language server. I'd love to hear your thoughts, feedback, and ideas.

if claude code is already running builds, tests, linting, git, etc. why is my editor doing all that too?

i mostly just want to browse files, inspect the codebase, make a few edits, and get out of the way.

that's the question that kept bothering me.

over the last few months i've noticed my workflow has changed completely.

i'm spending less time writing code inside the editor and more time talking to an AI agent in the terminal. it writes the implementation, runs the build, fixes lint errors, executes tests, commits changes, searches the project, and does most of the heavy lifting.

but my editor is still behaving like i'm writing every line myself.

it's indexing the project, spinning up language servers, running background processes, loading extensions, and consuming memory and battery for features i barely touch anymore.

at the same time, the terminal alone isn't enough.

i still want to inspect the project structure.

i want to quickly jump between files, review what the agent changed, understand how it's organizing the code, search through the project, and occasionally fix something by hand.

i don't want a browser tab pretending to be an editor.

i don't want another full IDE.

i just want something lightweight that complements the agent instead of competing with it.

that's why i started building Jelly.

the idea is simple: keep the editor focused on the things humans are still best at.

understanding architecture.

reviewing changes.

spotting when something feels off.

making small edits.

guiding the agent.

everything else can stay in the terminal where the agent already lives.

i'm not trying to replace VS Code, Cursor, or Zed. they're fantastic tools, and they make perfect sense if you're writing most of the code yourself.

but if your workflow has shifted toward AI agents, maybe the editor should evolve too.

maybe it doesn't need to do everything anymore.

maybe it just needs to stay fast, stay out of the way, and give you a clean view of your codebase.

that's the experiment i'm exploring with Jelly.

it's still very early, and i'm sure i'm wrong about a lot of things.

i'd genuinely love to hear how your workflow has changed.

if you're using claude code, codex, gemini cli, or other agentic tools, what do you actually need from your editor today?

repo: